---
title: "Level 3. Data Processing"
id: "1785"
type: "page"
slug: "data-processing"
published_at: "2026-09-20T21:35:45+00:00"
modified_at: "2026-09-20T21:43:08+00:00"
url: "https://xedant.com/agents/analytics/docs/templates/data-processing"
markdown_url: "https://xedant.com/agents/analytics/docs/templates/data-processing.md"
excerpt: "The third level is about turning raw files into data that answers questions. Join orders…"
---

# Level 3. Data Processing

[https://xedant.com/agents/analytics/docs/templates/data-processing.md](https://xedant.com/agents/analytics/docs/templates/data-processing.md)

The third level is about turning raw files into data that answers questions. Join orders to customers, put dates and names in order, gather rows into months, understand growth, and smooth out noise. Eight lessons — and every export stops being just a file.

Every lesson is a template with demo data: ten minutes to complete, one skill that stays. Open it in the Templates section of Analytics Agent and press “Use this template” to take the lesson on your own data.

### Joining Tables: Orders and Customers

**Shows:** on a shared key column, orders join to customers, and each order now shows its city and buyer; unmatched rows are not thrown away but counted honestly on the side — “4% unmatched” is more truthful than a “tidy 100%”.

**You get:** you learn where revenue by city comes from and who your best customers are — without asking anyone “please make a summary”.

### Dates and Texts in Order

**Shows:** one thing — one spelling: letter cases and synonyms unified, three date formats brought to one, money must be a number or the sum silently disappears. Every fix is a line in the log, so it can be rechecked.

**You get:** totals stop “vanishing” over little things like letter case or format — and you never have to explain why the numbers diverged.

### Grouping: Months and Categories

**Shows:** hundreds of rows compress into a dozen or two groups with one number per group — a “month × category” matrix shows peaks and quiet dips that raw rows never reveal. The total never changes: views change — the data never does.

**You get:** “group → count → chart” is the backbone of any report, and now it is in your hands.

### Growth: Day, Month and Year over Year

**Shows:** every percentage is a pair, “change + base”: compare with the same season last year, not with the neighboring stretch of time; month length is a base too (a 14.3% drop in February can be three missing days).

**You get:** you no longer conclude “we are falling!” from one percentage without a base — and stop scaring yourself with seasonality.

### Moving Average: Calming the Noise

**Shows:** daily jumps of ±13.5% turn into ±1.0%, and a real step change becomes a visible stair. The window is the dial: seven days dampens the weekly rhythm, twenty-eight erases the change itself — so keep both lines in view.

**You get:** you separate noise from trend in your own sales and stop jumping at every swing.

### Indexes and Buckets: Comparing the Unlike

**Shows:** an index gives lines of different scales a shared start (equal to 100), so you compare shape, not height; rank buckets split customers into quarters, and the top one carries 65% of the money — “small / medium / large” becomes countable.

**You get:** the “did it grow or not” comparison works even when the sums are dozens of times different.

### Asking Data in Plain Words

**Shows:** any analytics question is four moves: filter → sort → group → aggregate; both “the worst month” and “what brings in the most money” are built from them. Name the moves and you have asked a question data knows how to answer.

**You get:** you phrase questions to data in plain words — and get numbers back, not “just look at it yourself”.

### Facts and Dimensions: A First Data Model

**Shows:** a fact records an event — a purchase — and dimensions describe it: who, what, when; “revenue by city” is impossible from one orders table — you need a customer directory and a key. The two-way check: one number reached by two join paths must match.

**You get:** an understanding of why two reports about the same store show different revenue, and how to build data that does not argue with itself.

[← Templates library](/agents/analytics/docs/templates)
 · [Training course section](/agents/analytics/docs/templates#wp-block-themeisle-blocks-advanced-heading-aaat3)